21_gradient_boosting.ipynb

{
 "cells": [
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "`Gradient Boosting` is used to produce a good predictive model from an ensemble of weak models. For, datasets larger than 10_000 `Histogram Gradient Boosting` is more suitable"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd \n",
    "import numpy as np \n",
    "\n",
    "import matplotlib.pyplot as plt \n",
    "%matplotlib inline \n",
    "\n",
    "from sklearn import datasets, ensemble \n",
    "from sklearn.inspection import permutation_importance \n",
    "from sklearn.metrics import mean_squared_error \n",
    "from sklearn.model_selection import train_test_split "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "diabetes = datasets.load_diabetes()\n",
    "X, y = diabetes.data, diabetes.target"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Fitting with `GradientBoostingRegressor`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The Mean Squared Error (MSE) on test set: 3004.888\n"
     ]
    }
   ],
   "source": [
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.1, random_state=13)\n",
    "\n",
    "params = {\"n_estimators\": 500, \"max_depth\":4, \"min_samples_split\":5, \"learning_rate\":0.01, \"loss\":\"squared_error\"}\n",
    "\n",
    "reg_model = ensemble.GradientBoostingRegressor(**params)\n",
    "reg_model.fit(X_train, y_train)\n",
    "\n",
    "mse = mean_squared_error(y_test, reg_model.predict(X_test))\n",
    "print(\"The Mean Squared Error (MSE) on test set: {:.3f}\".format(mse))"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Plotting training Deviance"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_score = np.zeros((params[\"n_estimators\"],),dtype=np.float64)\n",
    "for i, y_pred in enumerate(reg_model.staged_predict(X_test)):\n",
    "    test_score[i] = mean_squared_error(y_test, y_pred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 432x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig = plt.figure(figsize=(6,6))\n",
    "plt.subplot(1,1,1)\n",
    "plt.title(\"Deviance\")\n",
    "plt.plot(np.arange(params[\"n_estimators\"]) +1, reg_model.train_score_, \"b-\", label=\"Training Set Deviance\")\n",
    "plt.plot(np.arange(params[\"n_estimators\"]) +1, test_score, \"r-\", label=\"Test Set Deviance\")\n",
    "plt.legend(loc = \"upper right\")\n",
    "plt.xlabel(\"Boosting Iterations\")\n",
    "plt.ylabel(\"Deviance\")\n",
    "fig.tight_layout()\n",
    "plt.show()"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Plotting `Feature Importance` and `Permutation Feature` <br>\n",
    "\n",
    "If we have predictive model $m$, dataset, $D$, ref_score $s$, the `permutation importance` algorithm is as: <br>\n",
    "- For each feature $j$ (column of $D$):<br>\n",
    "-- For each repetition $k$ in $1,..,K$:<br>\n",
    "-- -- Randomly shuffle column $j$ to generate a corrupted version of data - $D_{k,j}$ <br>\n",
    "-- -- Compute importance $i_j$ for feature $f_j$ defined as:<br>\n",
    "&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;$i_j = s - \\frac{1}{K}\\sum_{k=1}^Ks_{k,j}$"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 720x432 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "feature_importance = reg_model.feature_importances_ \n",
    "sorted_idx = np.argsort(feature_importance) \n",
    "\n",
    "pos = np.arange(sorted_idx.shape[0]) + 0.5 \n",
    "\n",
    "fig = plt.figure(figsize=(10,6))\n",
    "plt.subplot(1,2,1)\n",
    "plt.barh(pos, feature_importance[sorted_idx], align=\"center\")\n",
    "plt.yticks(pos, np.array(diabetes.feature_names)[sorted_idx])\n",
    "plt.title(\"Feature Importance (MDI)\")\n",
    "\n",
    "result = permutation_importance(reg_model, X_test, y_test, n_repeats=10, random_state=42, n_jobs=2)\n",
    "sorted_idx = result.importances_mean.argsort()\n",
    "plt.subplot(1,2,2)\n",
    "plt.boxplot(result.importances[sorted_idx].T, vert=False, labels=np.array(diabetes.feature_names)[sorted_idx])\n",
    "plt.title(\"Permutation Importance (test set)\")\n",
    "fig.tight_layout()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.9.9"
  },
  "orig_nbformat": 4
 },
 "nbformat": 4,
 "nbformat_minor": 2
}